A first-year turned 'Ilya's 30 ML papers' into a site nobody can source

What a first-year shipped, and why he built it to stop burning Claude credits

30papers.com renders "the reading list Ilya Sutskever gave John Carmack" as an in-browser gallery: thirty ML papers, each with arxiv links and notes, wrapped in animated backgrounds. The author showed up in the thread to explain himself:

"Author here. First year CS student at Trinity College Dublin. I built this because when I was getting into reading research papers I ended up burning a ton of my Claude usage asking questions other people have probably already asked." — the author, notmcrowley

The list's intellectual spine is its theory section. An ML reader unpacked why the Kolmogorov-complexity papers belong on it at all:

"Ilya argues that the reason why neural networks generalize … is because they are effectively finding a simple description of their training data, converging down onto the limit of the Kolmorogov complexity."

The list nobody can pin on Ilya, and why readers stopped caring

The front-page framing invites one obvious question, and the thread asked it bluntly:

"Someone posts on X, 'These are Ilya's 30 papers,' gives no source … Then someone vibe codes a barely usable website based on that, and it lands on the HN front page? Is this correct?" — a skeptical reader

The provenance did get traced, to a 2024 tweet with 876k views rather than to Ilya, by a reader who went looking for the origin. His verdict defused the argument:

"whether it was actually by Ilya or not is not worthy of debate. Many of them are widely recognized for being good pedagogical resources … and others are landmark papers … Krizhevsky et al. (2012) introduced AlexNet … Vaswani et al. (2017) introduced the Transformer."

Great papers, rough wrapper: broken LaTeX and too many animations

The canon survived scrutiny; the wrapper did not. The "beginner friendly format" promise took direct fire, and the rendering drew worse:

"all the LaTeX formulas are messed up. The subscripts and superscripts are all flattened … Did the author actually try to read any of the articles?" — a reader who opened the first paper and hit mangled math

What saved the reception was the author's response. Rather than argue, he shipped fixes: toggles to kill the movement and the backgrounds, plus an offer to add his own per-paper notes, with the honest caveat that he has "not been formally educated on ML or AI." For readers who want depth now, a writer who spent two years reviewing all ~23 of the actual papers pointed to his own archive at 12gramsofcarbon.com. The real takeaway isn't whose list it is: a canon this durable now has a first-year's front door, and several deeper ones behind it.